@PhilMiller
Copy link

Valgrind reports the following leak as AppStream is called from Debian's isenkramd package recommender for newly attached hardware:

==26466== 72,364 bytes in 4,495 blocks are definitely lost in loss record 5,532 of 5,538
==26466== at 0x4C2CB8F: malloc (vg_replace_malloc.c:299)
==26466== by 0x6D70858: g_malloc (gmem.c:99)
==26466== by 0x6D8A13E: g_strdup (gstrfuncs.c:363)
==26466== by 0x135FBF24: as_component_set_origin (as-component.c:1029)
==26466== by 0x13602E50: as_component_set_from_variant (as-component.c:5352)
==26466== by 0x1360602A: as_cache_file_read (as-pool.c:1758)
==26466== by 0x13606799: as_pool_load_cache_file (as-pool.c:1036)
==26466== by 0x13606E43: as_pool_load_collection_data (as-pool.c:659)
==26466== by 0x13606F3C: as_pool_load (as-pool.c:1004)
==26466== by 0x703AFCD: ffi_call_unix64 (in /usr/lib/x86_64-linux-gnu/libffi.so.6.0.4)
==26466== by 0x703A93E: ffi_call (in /usr/lib/x86_64-linux-gnu/libffi.so.6.0.4)
==26466== by 0x6674EA0: ??? (in /usr/lib/python2.7/dist-packages/gi/_gi.x86_64-linux-gnu.so)

This is the only remaining leak observed that grows with usage after my pull request #193 is applied. As noted there, I don't have the depth of knowledge in this code to directly fix it, so hopefully one of the maintainers can.